The expiration time for new job artifacts in CI/CD pipelines is now 30 days (GitLab default). Previously generated artifacts in already completed jobs will not be affected by the change. The latest artifacts for all jobs in the latest successful pipelines will be kept. More information: https://gitlab.lrz.de/help/user/admin_area/settings/continuous_integration.html#default-artifacts-expiration

parser.py 90.8 KB
Newer Older
1
"""parsers.py - parser combinators for for DHParser
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Copyright 2016  by Eckhart Arnold (arnold@badw.de)
                Bavarian Academy of Sciences an Humanities (badw.de)

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied.  See the License for the specific language governing
permissions and limitations under the License.
Eckhart Arnold's avatar
Eckhart Arnold committed
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

Module ``parsers.py`` contains a number of classes that together
make up parser combinators for left-recursive grammers. For each
element of the extended Backus-Naur-Form as well as for a regular
expression token a class is defined. The set of classes can be used to
define a parser for (ambiguous) left-recursive grammers.


References and Acknowledgements:

Dominikus Herzberg: Objekt-orientierte Parser-Kombinatoren in Python,
Blog-Post, September, 18th 2008 on denkspuren. gedanken, ideen,
anregungen und links rund um informatik-themen, URL:
http://denkspuren.blogspot.de/2008/09/objekt-orientierte-parser-kombinatoren.html

Dominikus Herzberg: Eine einfache Grammatik für LaTeX, Blog-Post,
September, 18th 2008 on denkspuren. gedanken, ideen, anregungen und
links rund um informatik-themen, URL:
http://denkspuren.blogspot.de/2008/09/eine-einfache-grammatik-fr-latex.html

Dominikus Herzberg: Uniform Syntax, Blog-Post, February, 27th 2007 on
denkspuren. gedanken, ideen, anregungen und links rund um
informatik-themen, URL:
http://denkspuren.blogspot.de/2007/02/uniform-syntax.html

Richard A. Frost, Rahmatullah Hafiz and Paul Callaghan: Parser
Combinators for Ambiguous Left-Recursive Grammars, in: P. Hudak and
D.S. Warren (Eds.): PADL 2008, LNCS 4902, pp. 167–181, Springer-Verlag
Berlin Heidelberg 2008.

47
48
49
50
Elizabeth Scott and Adrian Johnstone, GLL Parsing,
in: Electronic Notes in Theoretical Computer Science 253 (2010) 177–189,
http://dotat.at/tmp/gll.pdf

Eckhart Arnold's avatar
Eckhart Arnold committed
51
52
Juancarlo Añez: grako, a PEG parser generator in Python,
https://bitbucket.org/apalala/grako
53
54
55

Vegard Øye: General Parser Combinators in Racket, 2012,
https://epsil.github.io/gll/
56
57
"""

58
import collections
59
import copy
60
import html
61
import os
Eckhart Arnold's avatar
Eckhart Arnold committed
62
63
from functools import partial

64
from DHParser.error import Error, is_error, has_errors, linebreaks, line_col
65
66
67
68
from DHParser.stringview import StringView, EMPTY_STRING_VIEW
from DHParser.syntaxtree import Node, TransformationFunc, ParserBase, WHITESPACE_PTYPE, \
    TOKEN_PTYPE, ZOMBIE_PARSER
from DHParser.toolkit import is_logging, log_dir, logfile_basename, escape_re, sane_parser_name, \
di68kap's avatar
di68kap committed
69
70
71
    load_if_file, re, typing
from typing import Any, Callable, cast, Dict, List, Set, Tuple, Union, Optional

Eckhart Arnold's avatar
Eckhart Arnold committed
72

73
__all__ = ('PreprocessorFunc',
74
           'HistoryRecord',
Eckhart Arnold's avatar
Eckhart Arnold committed
75
           'Parser',
76
           'Grammar',
77
78
79
           'RX_PREPROCESSOR_TOKEN',
           'BEGIN_TOKEN',
           'END_TOKEN',
Eckhart Arnold's avatar
Eckhart Arnold committed
80
           'make_token',
81
82
           'nil_preprocessor',
           'PreprocessorToken',
Eckhart Arnold's avatar
Eckhart Arnold committed
83
84
85
86
           'RegExp',
           'RE',
           'Token',
           'mixin_comment',
87
88
89
           # 'UnaryOperator',
           # 'NaryOperator',
           'Synonym',
90
           'Option',
Eckhart Arnold's avatar
Eckhart Arnold committed
91
92
           'ZeroOrMore',
           'OneOrMore',
93
           'Series',
Eckhart Arnold's avatar
Eckhart Arnold committed
94
           'Alternative',
95
96
           'AllOf',
           'SomeOf',
97
           'Unordered',
Eckhart Arnold's avatar
Eckhart Arnold committed
98
99
100
101
           'Lookahead',
           'NegativeLookahead',
           'Lookbehind',
           'NegativeLookbehind',
102
103
104
           'last_value',
           'counterpart',
           'accumulate',
Eckhart Arnold's avatar
Eckhart Arnold committed
105
106
107
108
           'Capture',
           'Retrieve',
           'Pop',
           'Forward',
109
           'Compiler',
110
           'compile_source')
Eckhart Arnold's avatar
Eckhart Arnold committed
111

112

113
114
115
116
117
118
119
########################################################################
#
# Grammar and parsing infrastructure
#
########################################################################


120
PreprocessorFunc = Union[Callable[[str], str], partial]
121
122


123
124
125
LEFT_RECURSION_DEPTH = 8  # type: int
# because of python's recursion depth limit, this value ought not to be
# set too high. PyPy allows higher values than CPython
Eckhart Arnold's avatar
Eckhart Arnold committed
126
MAX_DROPOUTS = 3  # type: int
127
# stop trying to recover parsing after so many errors
128

129

130
131
132
class HistoryRecord:
    """
    Stores debugging information about one completed step in the
133
    parsing history.
134
135
136
137
138
139
140
141
142

    A parsing step is "completed" when the last one of a nested
    sequence of parser-calls returns. The call stack including
    the last parser call will be frozen in the ``HistoryRecord``-
    object. In addition a reference to the generated leaf node
    (if any) will be stored and the result status of the last
    parser call, which ist either MATCH, FAIL (i.e. no match)
    or ERROR.
    """
143
    __slots__ = ('call_stack', 'node', 'text', 'line_col')
144
145
146
147

    MATCH = "MATCH"
    ERROR = "ERROR"
    FAIL = "FAIL"
148
    Snapshot = collections.namedtuple('Snapshot', ['line', 'column', 'stack', 'status', 'text'])
eckhart's avatar
eckhart committed
149

di68kap's avatar
di68kap committed
150
151
    COLGROUP = '<colgroup>\n<col style="width:2%"/><col style="width:2%"/><col style="width:75"/>' \
               '<col style="width:6%"/><col style="width:15%"/>\n</colgroup>\n'
eckhart's avatar
eckhart committed
152
153
    HTML_LEAD_IN  = (
        '<html>\n<head>\n<meta charset="utf-8"/>\n<style>\n'
di68kap's avatar
di68kap committed
154
155
156
157
        'td.line, td.column {font-family:monospace;color:darkgrey}\n'
        'td.stack{font-family:monospace}\n'
        'td.status{font-family:monospace;font-weight:bold}\n'
        'td.text{font-family:monospace;color:darkblue}\n'
eckhart's avatar
eckhart committed
158
        'table{border-spacing: 0px; border: thin solid darkgrey; width:100%}\n'
di68kap's avatar
di68kap committed
159
160
161
162
163
        'td{border-right: thin solid grey; border-bottom: thin solid grey}\n'
        'span.delimiter{color:grey;}\nspan.match{color:darkgreen}\n'
        'span.fail{color:darkgrey}\nspan.error{color:red}\n'
        'span.matchstack{font-weight:bold;color:darkred}'
        '\n</style>\n</head>\n<body>\n<table>\n' + COLGROUP)
eckhart's avatar
eckhart committed
164
    HTML_LEAD_OUT = '\n</table>\n</body>\n</html>\n'
165

166
    def __init__(self, call_stack: List['Parser'], node: Node, text: StringView) -> None:
167
        # copy call stack, dropping uninformative Forward-Parsers
Eckhart Arnold's avatar
Eckhart Arnold committed
168
        self.call_stack = [p for p in call_stack if p.ptype != ":Forward"]  # type: List['Parser']
169
        self.node = node                # type: Node
170
        self.text = text                # type: StringView
171
172
        self.line_col = (1, 1)          # type: Tuple[int, int]
        if call_stack:
173
174
175
            grammar = call_stack[-1].grammar
            document = grammar.document__
            lbreaks = grammar.document_lbreaks__
176
            self.line_col = line_col(lbreaks, len(document) - len(text))
177
178

    def __str__(self):
179
180
181
        return '%4i, %2i:  %s;  %s;  "%s"' % self.as_tuple()

    def as_tuple(self) -> Snapshot:
eckhart's avatar
eckhart committed
182
183
        return self.Snapshot(self.line_col[0], self.line_col[1],
                             self.stack, self.status, self.excerpt)
184

eckhart's avatar
eckhart committed
185
    def as_html_tr(self) -> str:
eckhart's avatar
eckhart committed
186
187
188
189
190
191
        stack = html.escape(self.stack).replace(
            '-&gt;', '<span class="delimiter">&shy;-&gt;</span>')
        status = html.escape(self.status)
        excerpt = html.escape(self.excerpt)
        if status == self.MATCH:
            status = '<span class="match">' + status + '</span>'
di68kap's avatar
di68kap committed
192
193
194
195
196
197
198
199
200
201
202
203
204
            i = stack.rfind('-&gt;')
            chr = stack[i+12:i+13]
            while not chr.isidentifier() and i >= 0:
                i = stack.rfind('-&gt;', 0, i)
                chr = stack[i+12:i+13]
            if i >= 0:
                i += 12
                k = stack.find('<', i)
                if k < 0:
                    stack = stack[:i] + '<span class="matchstack">' + stack[i:]
                else:
                    stack = stack[:i] + '<span class="matchstack">' + stack[i:k] \
                            + '</span>' + stack[k:]
eckhart's avatar
eckhart committed
205
206
207
208
209
210
        elif status == self.FAIL:
            status = '<span class="fail">' + status + '</span>'
        else:
            stack += '<br/>\n' + status
            status = '<span class="error">ERROR</span>'
        tpl = self.Snapshot(str(self.line_col[0]), str(self.line_col[1]), stack, status, excerpt)
di68kap's avatar
di68kap committed
211
        # return ''.join(['<tr>'] + [('<td>%s</td>' % item) for item in tpl] + ['</tr>'])
eckhart's avatar
eckhart committed
212
        return ''.join(['<tr>'] + [('<td class="%s">%s</td>' % (cls, item))
eckhart's avatar
eckhart committed
213
                                   for cls, item in zip(tpl._fields, tpl)] + ['</tr>'])
214

215
    def err_msg(self) -> str:
216
217
        return self.ERROR + ": " + "; ".join(
            str(e) for e in (self.node._errors if self.node._errors else
218
                             self.node.collect_errors()[:2]))
219

220
    @property
221
    def stack(self) -> str:
222
        return "->".join((p.repr if p.ptype == ':RegExp' else p.name or p.ptype)
223
                         for p in self.call_stack)
224
225

    @property
226
    def status(self) -> str:
227
        return self.FAIL if self.node is None else \
228
229
            ('"%s"' % self.err_msg()) if self.node.error_flag else self.MATCH  # has_errors(self.node._errors)

eckhart's avatar
eckhart committed
230
231
232
    @property
    def excerpt(self):
        length = len(self.node) if self.node else len(self.text)
di68kap's avatar
di68kap committed
233
        excerpt = str(self.node)[:min(length, 20)] if self.node else self.text[:20]
eckhart's avatar
eckhart committed
234
235
236
237
238
        excerpt = excerpt.replace('\n', '\\n')
        if length > 20:
            excerpt += '...'
        return excerpt

239
240
241
242
    # @property
    # def extent(self) -> slice:
    #     return (slice(-self.remaining - len(self.node), -self.remaining) if self.node
    #             else slice(-self.remaining, None))
243
244

    @property
245
246
    def remaining(self) -> int:
        return len(self.text) - (len(self.node) if self.node else 0)
247

248
    @staticmethod
Eckhart Arnold's avatar
Eckhart Arnold committed
249
    def last_match(history: List['HistoryRecord']) -> Union['HistoryRecord', None]:
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
        """
        Returns the last match from the parsing-history.
        Args:
            history:  the parsing-history as a list of HistoryRecord objects

        Returns:
            the history record of the last match or none if either history is
            empty or no parser could match
        """
        for record in reversed(history):
            if record.status == HistoryRecord.MATCH:
                return record
        return None

    @staticmethod
Eckhart Arnold's avatar
Eckhart Arnold committed
265
    def most_advanced_match(history: List['HistoryRecord']) -> Union['HistoryRecord', None]:
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
        """
        Returns the closest-to-the-end-match from the parsing-history.
        Args:
            history:  the parsing-history as a list of HistoryRecord objects

        Returns:
            the history record of the closest-to-the-end-match or none if either history is
            empty or no parser could match
        """
        remaining = -1
        result = None
        for record in history:
            if (record.status == HistoryRecord.MATCH and
                    (record.remaining < remaining or remaining < 0)):
                result = record
                remaining = record.remaining
        return result


285
def add_parser_guard(parser_func):
286
287
    """
    Add a wrapper function to a parser functions (i.e. Parser.__call__ method)
288
    that takes care of memoizing, left recursion and, optionally, tracing
289
290
    (aka "history tracking") of parser calls. Returns the wrapped call.
    """
291
    def guarded_call(parser: 'Parser', text: StringView) -> Tuple[Optional[Node], StringView]:
292
        try:
293
            location = len(text)    # mind that location is always the distance to the end
294
295
            grammar = parser.grammar  # grammar may be 'None' for unconnected parsers!

296
297
            if grammar.last_rb__loc__ <= location:
                grammar.rollback_to__(location)
298

299
300
301
            # if location has already been visited by the current parser,
            # return saved result
            if location in parser.visited:
di68kap's avatar
di68kap committed
302
                return parser.visited[location]
303

Eckhart Arnold's avatar
Eckhart Arnold committed
304
305
            if grammar.history_tracking__:
                grammar.call_stack__.append(parser)
306
                grammar.moving_forward__ = True
Eckhart Arnold's avatar
Eckhart Arnold committed
307

308
309
310
311
312
313
            # break left recursion at the maximum allowed depth
            if grammar.left_recursion_handling__:
                if parser.recursion_counter.setdefault(location, 0) > LEFT_RECURSION_DEPTH:
                    grammar.recursion_locations__.add(location)
                    return None, text
                parser.recursion_counter[location] += 1
314
315
316
317

            # run original __call__ method
            node, rest = parser_func(parser, text)

318
319
320
            if grammar.left_recursion_handling__:
                parser.recursion_counter[location] -= 1

321
            if node is None:
322
                # retrieve an earlier match result (from left recursion) if it exists
323
324
                if location in grammar.recursion_locations__:
                    if location in parser.visited:
di68kap's avatar
di68kap committed
325
                        node, rest = parser.visited[location]
326
                        # TODO: maybe add a warning about occurrence of left-recursion here?
327
328
329
                    # don't overwrite any positive match (i.e. node not None) in the cache
                    # and don't add empty entries for parsers returning from left recursive calls!
                elif grammar.memoization__:
330
                    # otherwise also cache None-results
di68kap's avatar
di68kap committed
331
                    parser.visited[location] = (None, rest)
Eckhart Arnold's avatar
Eckhart Arnold committed
332
333
            elif (grammar.last_rb__loc__ > location
                  and (grammar.memoization__ or location in grammar.recursion_locations__)):
334
335
336
337
                # - variable manipulating parsers will not be entered into the cache,
                #   because caching would interfere with changes of variable state
                # - in case of left recursion, the first recursive step that
                #   matches will store its result in the cache
di68kap's avatar
di68kap committed
338
                parser.visited[location] = (node, rest)
339

340
            if grammar.history_tracking__:
341
                # don't track returning parsers except in case an error has occurred
342
                remaining = len(rest)
343
                if grammar.moving_forward__ or (node and node.error_flag):  # node._errors
344
                    record = HistoryRecord(grammar.call_stack__, node, text)
345
                    grammar.history__.append(record)
346
                    # print(record.stack, record.status, rest[:20].replace('\n', '|'))
347
                grammar.moving_forward__ = False
348
                grammar.call_stack__.pop()
349

350
        except RecursionError:
351
            node = Node(None, str(text[:min(10, max(1, text.find("\n")))]) + " ...")
352
353
            node.add_error("maximum recursion depth of parser reached; "
                           "potentially due to too many errors!")
354
            rest = EMPTY_STRING_VIEW
355
356
357
358
359
360

        return node, rest

    return guarded_call


361
class Parser(ParserBase):
362
363
364
365
    """
    (Abstract) Base class for Parser combinator parsers. Any parser
    object that is actually used for parsing (i.e. no mock parsers)
    should should be derived from this class.
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382

    Since parsers can contain other parsers (see classes UnaryOperator
    and NaryOperator) they form a cyclical directed graph. A root
    parser is a parser from which all other parsers can be reached.
    Usually, there is one root parser which serves as the starting
    point of the parsing process. When speaking of "the root parser"
    it is this root parser object that is meant.

    There are two different types of parsers:

    1. *Named parsers* for which a name is set in field parser.name.
       The results produced by these parsers can later be retrieved in
       the AST by the parser name.

    2. *Anonymous parsers* where the name-field just contains the empty
       string. AST-transformation of Anonymous parsers can be hooked
       only to their class name, and not to the individual parser.
Eckhart Arnold's avatar
Eckhart Arnold committed
383
384

    Parser objects are callable and parsing is done by calling a parser
Eckhart Arnold's avatar
Eckhart Arnold committed
385
386
387
388
389
390
    object with the text to parse.

    If the parser matches it returns a tuple consisting of a node
    representing the root of the concrete syntax tree resulting from the
    match as well as the substring `text[i:]` where i is the length of
    matched text (which can be zero in the case of parsers like
391
    `ZeroOrMore` or `Option`). If `i > 0` then the parser has "moved
Eckhart Arnold's avatar
Eckhart Arnold committed
392
393
394
395
396
397
    forward".

    If the parser does not match it returns `(None, text). **Note** that
    this is not the same as an empty match `("", text)`. Any empty match
    can for example be returned by the `ZeroOrMore`-parser in case the
    contained parser is repeated zero times.
398

399
    Attributes and Properties:
400
        visited:  Mapping of places this parser has already been to
di68kap's avatar
di68kap committed
401
                during the current parsing process onto the results the
402
403
404
405
406
407
408
409
                parser returned at the respective place. This dictionary
                is used to implement memoizing.

        recursion_counter:  Mapping of places to how often the parser
                has already been called recursively at this place. This
                is needed to implement left recursion. The number of
                calls becomes irrelevant once a resault has been memoized.

410
        cycle_detection:  The apply()-method uses this variable to make
411
412
413
                sure that one and the same function will not be applied
                (recursively) a second time, if it has already been
                applied to this parser.
414
415
416

        grammar:  A reference to the Grammar object to which the parser
                is attached.
417
418
    """

419
420
    ApplyFunc = Callable[['Parser'], None]

421
422
    def __init__(self, name: str = '') -> None:
        # assert isinstance(name, str), str(name)
Eckhart Arnold's avatar
Eckhart Arnold committed
423
        super(Parser, self).__init__(name)
424
        self._grammar = None  # type: 'Grammar'
425
426
        self.reset()

427
428
        # add "aspect oriented" wrapper around parser calls
        # for memoizing, left recursion and tracing
Eckhart Arnold's avatar
Eckhart Arnold committed
429

430
431
432
433
434
435
436
        guarded_parser_call = add_parser_guard(self.__class__.__call__)
        # The following check is necessary for classes that don't override
        # the __call__() method, because in these cases the non-overridden
        # __call__()-method would be substituted a second time!
        if self.__class__.__call__.__code__ != guarded_parser_call.__code__:
            self.__class__.__call__ = guarded_parser_call

Eckhart Arnold's avatar
Eckhart Arnold committed
437
    def __deepcopy__(self, memo):
Eckhart Arnold's avatar
Eckhart Arnold committed
438
439
440
441
442
443
        """Deepcopy method of the parser. Upon instantiation of a Grammar-
        object, parsers will be deep-copied to the Grammar object. If a
        derived parser-class changes the signature of the constructor,
        `__deepcopy__`-method must be replaced (i.e. overridden without
        calling the same method from the superclass) by the derived class.
        """
444
445
        return self.__class__(self.name)

446
    def reset(self):
Eckhart Arnold's avatar
Eckhart Arnold committed
447
448
449
        """Initializes or resets any parser variables. If overwritten,
        the `reset()`-method of the parent class must be called from the
        `reset()`-method of the derived class."""
450
        self.visited = dict()            # type: Dict[int, Tuple[Optional[Node], StringView]]
451
        self.recursion_counter = dict()  # type: Dict[int, int]
452
        self.cycle_detection = set()     # type: Set[Callable]
453

454
    def __call__(self, text: StringView) -> Tuple[Optional[Node], StringView]:
Eckhart Arnold's avatar
Eckhart Arnold committed
455
456
457
        """Applies the parser to the given `text` and returns a node with
        the results or None as well as the text at the position right behind
        the matching string."""
458
459
        return None, text  # default behaviour: don't match

460
    def __add__(self, other: 'Parser') -> 'Series':
461
462
        """The + operator generates a series-parser that applies two
        parsers in sequence."""
463
        return Series(self, other)
464

465
    def __or__(self, other: 'Parser') -> 'Alternative':
466
467
468
        """The | operator generates an alternative parser that applies
        the first parser and, if that does not match, the second parser.
        """
469
470
        return Alternative(self, other)

471
    @property
472
    def grammar(self) -> 'Grammar':
473
474
475
        return self._grammar

    @grammar.setter
476
    def grammar(self, grammar: 'Grammar'):
Eckhart Arnold's avatar
Eckhart Arnold committed
477
478
479
480
481
482
        if self._grammar is None:
            self._grammar = grammar
            self._grammar_assigned_notifier()
        else:
            assert self._grammar == grammar, \
                "Parser has already been assigned to a different Grammar object!"
483
484

    def _grammar_assigned_notifier(self):
485
486
        """A function that notifies the parser object that it has been
        assigned to a grammar."""
487
488
        pass

489
    def apply(self, func: ApplyFunc) -> bool:
490
491
        """
        Applies function `func(parser)` recursively to this parser and all
Eckhart Arnold's avatar
Eckhart Arnold committed
492
        descendant parsers if any exist. The same function can never
493
        be applied twice between calls of the ``reset()``-method!
Eckhart Arnold's avatar
Eckhart Arnold committed
494
495
        Returns `True`, if function has been applied, `False` if function
        had been applied earlier already and thus has not been applied again.
496
497
498
499
        """
        if func in self.cycle_detection:
            return False
        else:
Eckhart Arnold's avatar
Eckhart Arnold committed
500
501
            assert not self.visited, "No calls to Parser.apply() during or " \
                                     "after ongoing parsing process. (Call Parser.reset() first.)"
502
503
504
505
506
            self.cycle_detection.add(func)
            func(self)
            return True


507
def mixin_comment(whitespace: str, comment: str) -> str:
508
509
    """
    Returns a regular expression that merges comment and whitespace
510
511
512
513
514
515
516
517
518
519
520
    regexps. Thus comments cann occur whereever whitespace is allowed
    and will be skipped just as implicit whitespace.

    Note, that because this works on the level of regular expressions,
    nesting comments is not possible. It also makes it much harder to
    use directives inside comments (which isn't recommended, anyway).
    """
    wspc = '(?:' + whitespace + '(?:' + comment + whitespace + ')*)'
    return wspc


521
class Grammar:
eckhart's avatar
eckhart committed
522
    r"""
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
    Class Grammar directs the parsing process and stores global state
    information of the parsers, i.e. state information that is shared
    accross parsers.

    Grammars are basically collections of parser objects, which are
    connected to an instance object of class Grammar. There exist two
    ways of connecting parsers to grammar objects: Either by passing
    the root parser object to the constructor of a Grammar object
    ("direct instantiation"), or by assigning the root parser to the
    class variable "root__" of a descendant class of class Grammar.

    Example for direct instantian of a grammar:

        >>> number = RE('\d+') + RE('\.') + RE('\d+') | RE('\d+')
        >>> number_parser = Grammar(number)
Eckhart Arnold's avatar
Eckhart Arnold committed
538
        >>> number_parser("3.1416").content()
539
540
        '3.1416'

Eckhart Arnold's avatar
Eckhart Arnold committed
541
    Collecting the parsers that define a grammar in a descendant class of
542
543
544
545
546
547
548
    class Grammar and assigning the named parsers to class variables
    rather than global variables has several advantages:

    1. It keeps the namespace clean.

    2. The parser names of named parsers do not need to be passed to the
       constructor of the Parser object explicitly, but it suffices to
Eckhart Arnold's avatar
Eckhart Arnold committed
549
550
       assign them to class variables, which results in better
       readability of the Python code.
551

Eckhart Arnold's avatar
Eckhart Arnold committed
552
553
554
    3. The parsers in the class do not necessarily need to be connected
       to one single root parser, which is helpful for testing and
       building up a parser successively of several components.
555
556
557
558
559
560

    As a consequence, though, it is highly recommended that a Grammar
    class should not define any other variables or methods with names
    that are legal parser names. A name ending with a double
    underscore '__' is *not* a legal parser name and can safely be
    used.
561
562

    Example:
563

564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
        class Arithmetic(Grammar):
            # special fields for implicit whitespace and comment configuration
            COMMENT__ = r'#.*(?:\n|$)'  # Python style comments
            wspR__ = mixin_comment(whitespace=r'[\t ]*', comment=COMMENT__)

            # parsers
            expression = Forward()
            INTEGER = RE('\\d+')
            factor = INTEGER | Token("(") + expression + Token(")")
            term = factor + ZeroOrMore((Token("*") | Token("/")) + factor)
            expression.set(term + ZeroOrMore((Token("+") | Token("-")) + term))
            root__ = expression

    Upon instantiation the parser objects are deep-copied to the
    Grammar object and assigned to object variables of the same name.
    Any parser that is directly assigned to a class variable is a
    'named' parser and its field `parser.name` contains the variable
    name after instantiation of the Grammar class. All other parsers,
582
583
584
    i.e. parsers that are defined within a `named` parser, remain
    "anonymous parsers" where `parser.name` is the empty string, unless
    a name has been passed explicitly upon instantiation.
585
586
587
588
589
590
591
592
593
594
595
    If one and the same parser is assigned to several class variables
    such as, for example the parser `expression` in the example above,
    the first name sticks.

    Grammar objects are callable. Calling a grammar object with a UTF-8
    encoded document, initiates the parsing of the document with the
    root parser. The return value is the concrete syntax tree. Grammar
    objects can be reused (i.e. called again) after parsing. Thus, it
    is not necessary to instantiate more than one Grammar object per
    thread.

596
    Grammar classes contain a few special class fields for implicit
597
598
    whitespace and comments that should be overwritten, if the defaults
    (no comments, horizontal right aligned whitespace) don't fit:
599
600
    COMMENT__:  regular expression string for matching comments
    WSP__:   regular expression for whitespace and comments
Eckhart Arnold's avatar
Eckhart Arnold committed
601

602
603
    wspL__:  regular expression string for left aligned whitespace,
             which either equals WSP__ or is empty.
Eckhart Arnold's avatar
Eckhart Arnold committed
604

605
606
    wspR__:  regular expression string for right aligned whitespace,
             which either equals WSP__ or is empty.
Eckhart Arnold's avatar
Eckhart Arnold committed
607

608
609
610
611
    root__:  The root parser of the grammar. Theoretically, all parsers of the
             grammar should be reachable by the root parser. However, for testing
             of yet incomplete grammars class Grammar does not assume that this
             is the case.
Eckhart Arnold's avatar
Eckhart Arnold committed
612

613
614
615
616
617
618
619
620
621
    parser_initializiation__:  Before the parser class (!) has been initialized,
             which happens upon the first time it is instantiated (see doctring for
             method `_assign_parser_names()` for an explanation), this class
             field contains a value other than "done". A value of "done" indicates
             that the class has already been initialized.

    Attributes:
        all_parsers__:  A set of all parsers connected to this grammar object

Eckhart Arnold's avatar
Eckhart Arnold committed
622
        history_tracking__:  A flag indicating that the parsing history shall
623
624
625
626
627
628
629
                be tracked

        wsp_left_parser__:  A parser for the default left-adjacent-whitespace
                or the zombie-parser (see `syntaxtree.ZOMBIE_PARSER`) if the
                default is empty. The default whitespace will be used by parsers
                `Token` and, if no other parsers are passed to its constructor,
                by parser `RE'.
Eckhart Arnold's avatar
Eckhart Arnold committed
630

631
632
633
634
635
636
637
638
639
640
641
        wsp_right_parser__: The same for the default right-adjacent-whitespace.
                Both wsp_left_parser__ and wsp_right_parser__ merely serve the
                purpose to avoid having to specify the default whitespace
                explicitly every time an `RE`-parser-object is created.

        _dirty_flag__:  A flag indicating that the Grammar has been called at
                least once so that the parsing-variables need to be reset
                when it is called again.

        document__:  the text that has most recently been parsed or that is
                currently being parsed.
Eckhart Arnold's avatar
Eckhart Arnold committed
642

643
644
645
646
647
        document_lbreaks__:  list of linebreaks within the document, starting
                with -1 and ending with EOF. This helps generating line
                and column number for history recording and will only be
                initialized if `history_tracking__` is true.

648
649
        _reversed__:  the same text in reverse order - needed by the `Lookbehind'-
                parsers.
Eckhart Arnold's avatar
Eckhart Arnold committed
650

651
652
653
        variables__:  A mapping for variable names to a stack of their respective
                string values - needed by the `Capture`-, `Retrieve`- and `Pop`-
                parsers.
Eckhart Arnold's avatar
Eckhart Arnold committed
654

655
656
657
658
659
        rollback__:  A list of tuples (location, rollback-function) that are
                deposited by the `Capture`- and `Pop`-parsers. If the parsing
                process reaches a dead end then all rollback-functions up to
                the point to which it retreats will be called and the state
                of the variable stack restored accordingly.
Eckhart Arnold's avatar
Eckhart Arnold committed
660

661
662
663
664
665
666
667
668
        last_rb__loc__:  The last, i.e. most advanced location in the text
                where a variable changing operation occurred. If the parser
                backtracks to a location at or before `last_rb__loc__` (which,
                since locations are counted from the reverse, means:
                `location >= last_rb__loc__`) then a rollback of all variable
                changing operations is necessary that occurred after the
                location to which the parser backtracks. This is done by
                calling method `.rollback_to__(location)`.
Eckhart Arnold's avatar
Eckhart Arnold committed
669

670
671
672
673
        call_stack__:  A stack of all parsers that have been called. This
                is required for recording the parser history (for debugging)
                and, eventually, i.e. one day in the future, for tracing through
                the parsing process.
Eckhart Arnold's avatar
Eckhart Arnold committed
674

675
676
677
        history__:  A list of parser-call-stacks. A parser-call-stack is
                appended to the list each time a parser either matches, fails
                or if a parser-error occurs.
Eckhart Arnold's avatar
Eckhart Arnold committed
678

679
680
681
682
683
        moving_forward__: This flag indicates that the parsing process is currently
                moving forward . It is needed to reduce noise in history recording
                and should not be considered as having a valid value if history
                recording is turned off! (See `add_parser_guard` and its local
                function `guarded_call`)
Eckhart Arnold's avatar
Eckhart Arnold committed
684

685
686
687
688
689
        recursion_locations__:  Stores the locations where left recursion was
                detected. Needed to provide minimal memoization for the left
                recursion detection algorithm, but, strictly speaking, superfluous
                if full memoization is enabled. (See `add_parser_guard` and its
                local function `guarded_call`)
Eckhart Arnold's avatar
Eckhart Arnold committed
690

691
692
693
        memoization__:  Turns full memoization on or off. Turning memoization off
                results in less memory usage and sometimes reduced parsing time.
                In some situations it may drastically increase parsing time, so
694
695
696
697
698
                it is safer to leave it on. (Default: on)

        left_recursion_handling__:  Turns left recursion handling on or off.
                If turned off, a recursion error will result in case of left
                recursion.
699
    """
700
    root__ = ZOMBIE_PARSER  # type: ParserBase
701
    # root__ must be overwritten with the root-parser by grammar subclass
Eckhart Arnold's avatar
Eckhart Arnold committed
702
    parser_initialization__ = "pending"  # type: str
703
    # some default values
Eckhart Arnold's avatar
Eckhart Arnold committed
704
705
706
707
    COMMENT__ = r''  # type: str  # r'#.*(?:\n|$)'
    WSP__ = mixin_comment(whitespace=r'[\t ]*', comment=COMMENT__)  # type: str
    wspL__ = ''     # type: str
    wspR__ = WSP__  # type: str
708

709

710
    @classmethod
711
    def _assign_parser_names__(cls):
712
713
        """
        Initializes the `parser.name` fields of those
714
715
        Parser objects that are directly assigned to a class field with
        the field's name, e.g.
716
            class Grammar(Grammar):
717
718
719
                ...
                symbol = RE('(?!\\d)\\w+')
        After the call of this method symbol.name == "symbol"
720
721
722
723
        holds. Names assigned via the ``name``-parameter of the
        constructor will not be overwritten. Parser names starting or
        ending with a double underscore like ``root__`` will be
        ignored. See ``toolkit.sane_parser_name()``
724
725
726
727

        This is done only once, upon the first instantiation of the
        grammar class!

728
        Attention: If there exists more than one reference to the same
eckhart's avatar
eckhart committed
729
        parser, only the first one will be chosen for python versions
730
731
        greater or equal 3.6.  For python version <= 3.5 an arbitrarily
        selected reference will be chosen. See PEP 520
eckhart's avatar
eckhart committed
732
        (www.python.org/dev/peps/pep-0520/) for an explanation of why.
733
        """
734
735
736
737
738
        if cls.parser_initialization__ != "done":
            cdict = cls.__dict__
            for entry, parser in cdict.items():
                if isinstance(parser, Parser) and sane_parser_name(entry):
                    if not parser.name:
739
                        parser._name = entry
740
                    if isinstance(parser, Forward) and (not parser.parser._name):
741
                        parser.parser._name = entry
742
            cls.parser_initialization__ = "done"
743

744

eckhart's avatar
eckhart committed
745
    def __init__(self, root: Parser = None) -> None:
Eckhart Arnold's avatar
Eckhart Arnold committed
746
747
        # if not hasattr(self.__class__, 'parser_initialization__'):
        #     self.__class__.parser_initialization__ = "pending"
748
749
750
751
        # if not hasattr(self.__class__, 'wspL__'):
        #     self.wspL__ = ''
        # if not hasattr(self.__class__, 'wspR__'):
        #     self.wspR__ = ''
752
        self.all_parsers__ = set()             # type: Set[ParserBase]
753
754
755
        self._dirty_flag__ = False             # type: bool
        self.history_tracking__ = False        # type: bool
        self.memoization__ = True              # type: bool
756
        self.left_recursion_handling__ = True  # type: bool
757
        self._reset__()
758

Eckhart Arnold's avatar
Eckhart Arnold committed
759
        # prepare parsers in the class, first
760
        self._assign_parser_names__()
761

Eckhart Arnold's avatar
Eckhart Arnold committed
762
763
764
765
        # then deep-copy the parser tree from class to instance;
        # parsers not connected to the root object will be copied later
        # on demand (see Grammar.__getitem__()). Usually, the need to
        # do so only arises during testing.
Eckhart Arnold's avatar
Eckhart Arnold committed
766
        self.root__ = copy.deepcopy(root) if root else copy.deepcopy(self.__class__.root__)
767

768
        if self.wspL__:
Eckhart Arnold's avatar
Eckhart Arnold committed
769
            self.wsp_left_parser__ = Whitespace(self.wspL__)  # type: ParserBase
770
            self.wsp_left_parser__.grammar = self
771
            self.all_parsers__.add(self.wsp_left_parser__)  # don't you forget about me...
772
773
774
        else:
            self.wsp_left_parser__ = ZOMBIE_PARSER
        if self.wspR__:
Eckhart Arnold's avatar
Eckhart Arnold committed
775
            self.wsp_right_parser__ = Whitespace(self.wspR__)  # type: ParserBase
776
            self.wsp_right_parser__.grammar = self
777
            self.all_parsers__.add(self.wsp_right_parser__)  # don't you forget about me...
778
779
        else:
            self.wsp_right_parser__ = ZOMBIE_PARSER
780
        self.root__.apply(self._add_parser__)
781

782

783
    def __getitem__(self, key):
784
785
786
        try:
            return self.__dict__[key]
        except KeyError:
787
788
            parser_template = getattr(self, key, None)
            if parser_template:
Eckhart Arnold's avatar
Eckhart Arnold committed
789
                # add parser to grammar object on the fly...
790
                parser = copy.deepcopy(parser_template)
791
                parser.apply(self._add_parser__)
792
                # assert self[key] == parser
Eckhart Arnold's avatar
Eckhart Arnold committed
793
                return self[key]
794
            raise KeyError('Unknown parser "%s" !' % key)
795

796

797
    def _reset__(self):
798
799
        self.document__ = EMPTY_STRING_VIEW   # type: StringView
        self._reversed__ = EMPTY_STRING_VIEW  # type: StringView
800
        self.document_lbreaks__ = []          # type: List[int]
801
        # variables stored and recalled by Capture and Retrieve parsers
802
803
804
        self.variables__ = dict()             # type: Dict[str, List[str]]
        self.rollback__ = []                  # type: List[Tuple[int, Callable]]
        self.last_rb__loc__ = -1              # type: int
805
        # support for call stack tracing
806
        self.call_stack__ = []                # type: List[Parser]
807
        # snapshots of call stacks
808
        self.history__ = []                   # type: List[HistoryRecord]
809
        # also needed for call stack tracing
810
811
        self.moving_forward__ = False         # type: bool
        self.recursion_locations__ = set()    # type: Set[int]
812

813

Eckhart Arnold's avatar
Eckhart Arnold committed
814
    @property
Eckhart Arnold's avatar
Eckhart Arnold committed
815
    def reversed__(self) -> StringView:
eckhart's avatar
eckhart committed
816
817
818
819
820
        """
        Returns a reversed version of the currently parsed document. As
        about the only case where this is needed is the Lookbehind-parser,
        this is done lazily.
        """
Eckhart Arnold's avatar
Eckhart Arnold committed
821
        if not self._reversed__:
822
            self._reversed__ = StringView(self.document__.text[::-1])
Eckhart Arnold's avatar
Eckhart Arnold committed
823
824
825
        return self._reversed__


826
    def _add_parser__(self, parser: Parser) -> None:
827
828
        """
        Adds the particular copy of the parser object to this
829
        particular instance of Grammar.
830
        """
831
        if parser.name:
Eckhart Arnold's avatar
Eckhart Arnold committed
832
833
834
            # prevent overwriting instance variables or parsers of a different class
            assert parser.name not in self.__dict__ or \
                   isinstance(self.__dict__[parser.name], parser.__class__), \
835
836
                ('Cannot add parser "%s" because a field with the same name '
                 'already exists in grammar object!' % parser.name)
837
            setattr(self, parser.name, parser)
838
        self.all_parsers__.add(parser)
839
840
        parser.grammar = self

841

Eckhart Arnold's avatar
Eckhart Arnold committed
842
    def __call__(self, document: str, start_parser="root__") -> Node:
843
844
        """
        Parses a document with with parser-combinators.
845
846
847

        Args:
            document (str): The source text to be parsed.
848
849
850
            start_parser (str): The name of the parser with which to
                start. This is useful for testing particular parsers
                (i.e. particular parts of the EBNF-Grammar.)
851
852
853
        Returns:
            Node: The root node ot the parse tree.
        """
854
        # assert isinstance(document, str), type(document)
855
856
        if self.root__ is None:
            raise NotImplementedError()
857
        if self._dirty_flag__:
858
            self._reset__()
859
            for parser in self.all_parsers__:
860
861
                parser.reset()
        else:
862
            self._dirty_flag__ = True
863
        self.history_tracking__ = is_logging()
864
        self.document__ = StringView(document)
865
        self.document_lbreaks__ = linebreaks(document) if self.history_tracking__ else []
866
        self.last_rb__loc__ = len(self.document__) + 1  # rollback location
Eckhart Arnold's avatar
Eckhart Arnold committed
867
        parser = self[start_parser] if isinstance(start_parser, str) else start_parser
868
869
        assert parser.grammar == self, "Cannot run parsers from a different grammar object!" \
                                       " %s vs. %s" % (str(self), str(parser.grammar))
870
        result = None  # type: Optional[Node]
871
        stitches = []  # type: List[Node]
872
        rest = self.document__
873
        if not rest:
eckhart's avatar
eckhart committed
874
            result, _ = parser(rest)
di68kap's avatar
di68kap committed
875
876
877
            if result is None:
                result = Node(None, '')
                result.add_error('Parser "%s" did not match empty document.' % str(parser))
878
879
880
881
882
883
        while rest and len(stitches) < MAX_DROPOUTS:
            result, rest = parser(rest)
            if rest:
                fwd = rest.find("\n") + 1 or len(rest)
                skip, rest = rest[:fwd], rest[fwd:]
                if result is None:
884
885
886
887
                    error_msg = 'Parser did not match! Invalid source file?' \
                                '\n    Most advanced: %s\n    Last match:    %s;' % \
                                (str(HistoryRecord.most_advanced_match(self.history__)),
                                 str(HistoryRecord.last_match(self.history__)))
888
889
890
                else:
                    stitches.append(result)
                    error_msg = "Parser stopped before end" + \
891
892
                                (("! trying to recover" +
                                  (" but stopping history recording at this point."
893
                                   if self.history_tracking__ else "..."))
894
895
896
897
                                 if len(stitches) < MAX_DROPOUTS
                                 else " too often! Terminating parser.")
                stitches.append(Node(None, skip))
                stitches[-1].add_error(error_msg)
898
                if self.history_tracking__:
899
                    # some parsers may have matched and left history records with nodes != None.
900
                    # Because these are not connected to the stitched root node, their pos-
901
902
                    # properties will not be initialized by setting the root node's pos property
                    # to zero. Therefore, their pos properties need to be initialized here
903
                    for record in self.history__:
904
905
                        if record.node and record.node._pos < 0:
                            record.node.pos = 0
906
                    record = HistoryRecord(self.call_stack__.copy(), stitches[-1], rest)
907
                    self.history__.append(record)
908
909
                    # stop history tracking when parser returned too early
                    self.history_tracking__ = False
910
911
912
913
        if stitches:
            if rest:
                stitches.append(Node(None, rest))
            result = Node(None, tuple(stitches))
914
        if any(self.variables__.values()):
Eckhart Arnold's avatar
Eckhart Arnold committed
915
916
            error_str = "Capture-retrieve-stack not empty after end of parsing: " + \
                            str(self.variables__)
917
918
919
920
921
922
923
924
925
926
            if result:
                if result.children:
                    # add another child node at the end to ensure that the position
                    # of the error will be the end of the text. Otherwise, the error
                    # message above ("...after end of parsing") would appear illogical.
                    error_node = Node(ZOMBIE_PARSER, '')
                    error_node.add_error(error_str)
                    result.result = result.children + (error_node,)
                else:
                    result.add_error(error_str)
927
        result.pos = 0  # calculate all positions
928
        # result.collect_errors(self.document__)
929
930
        return result

931

932
    def push_rollback__(self, location, func):
933
934
        """
        Adds a rollback function that either removes or re-adds
935
936
937
938
        values on the variable stack (`self.variables`) that have been
        added (or removed) by Capture or Pop Parsers, the results of
        which have been dismissed.
        """
939
940
941
        self.rollback__.append((location, func))
        self.last_rb__loc__ = location

942

943
    def rollback_to__(self, location):
944
945
        """
        Rolls back the variable stacks (`self.variables`) to its
946
947
        state at an earlier location in the parsed document.
        """
948
        while self.rollback__ and self.rollback__[-1][0] <= location:
949
            _, rollback_func = self.rollback__.pop()
950
951
952
953
            # assert not loc > self.last_rb__loc__, \
            #     "Rollback confusion: line %i, col %i < line %i, col %i" % \
            #     (*line_col(self.document__, len(self.document__) - loc),
            #      *line_col(self.document__, len(self.document__) - self.last_rb__loc__))
954
955
956
957
            rollback_func()
        self.last_rb__loc__ == self.rollback__[-1][0] if self.rollback__ \
            else (len(self.document__) + 1)

958

di68kap's avatar
di68kap committed
959
    def log_parsing_history__(self, log_file_name: str = '', html: bool=True) -> None:
960
961
        """
        Writes a log of the parsing history of the most recently parsed
eckhart's avatar
eckhart committed
962
        document.
963
964
        """
        def write_log(history, log_name):
eckhart's avatar
eckhart committed
965
966
            htm = '.html' if html else ''
            path = os.path.join(log_dir(), log_name + "_parser.log" + htm)
Eckhart Arnold's avatar
Eckhart Arnold committed
967
968
969
            if os.path.exists(path):
                os.remove(path)
                print('WARNING: Log-file "%s" already existed and was deleted.' % path)
970
971
            if history:
                with open(path, "w", encoding="utf-8") as f:
eckhart's avatar
eckhart committed
972
973
974
975
976
977
                    if html:
                        f.write(HistoryRecord.HTML_LEAD_IN)
                        f.write("\n".join(history))
                        f.write(HistoryRecord.HTML_LEAD_OUT)
                    else:
                        f.write("\n".join(history))
978

di68kap's avatar
di68kap committed
979
980
981
982
983
984
985
986
        def append_line(log, line):
            """Appends a line to a list of HTML table rows. Starts a new
            table every 100 rows to allow browser to speed up rendering.
            Does this really work...?"""
            log.append(line)
            if html and len(log) % 100 == 0:
                log.append('\n</table>\n<table>\n' + HistoryRecord.COLGROUP)

987
988
989
990
991
992
        if is_logging():
            assert self.history__, \
                "Parser did not yet run or logging was turned off when running parser!"
            if not log_file_name:
                name = self.__class__.__name__
                log_file_name = name[:-7] if name.lower().endswith('grammar') else name